January 21, 2026

Prisma ORM 7.3.0: Faster queries, safer data, smoother upgrades

Prisma ORM 7.3.0 focuses on a theme we’ve been steadily pushing since 7.0: performance improvements without sacrificing reliability. This release introduces new ways to tune query compilation, reduces overhead for raw queries, includes important adapter updates, and ships a few correctness fixes that make upgrading and day-to-day usage smoother.

Prisma ORM 7.3.0: Faster queries, safer data, smoother upgrades

Choose the right query compiler for your app

One of the biggest changes in 7.3.0 is a new compilerBuild option in your Prisma Client generator. It lets you choose between two query compiler builds:

  • fast: optimized for speed (default), with a larger footprint
  • small: optimized for bundle size, trading off some performance

This is the first step in a larger set of performance work we’re continuing to roll out across the ORM.

Raw queries now skip unnecessary overhead

If you’re using raw queries ($executeRaw, $queryRaw), 7.3.0 reduces the work Prisma has to do before sending your query to the database. Raw queries can now bypass the query compiler and interpreter path and go straight to the driver adapter, removing extra layers and trimming latency in raw-query heavy workloads.

Adapter and platform updates

This release includes a couple of practical ecosystem improvements:

  • MSSQL adapter updated: @prisma/adapter-mssql now uses MSSQL v12.2.0, thanks to community contributor Jay-Lokhande.
  • SQLite stability fix for better-sqlite3: an underlying SQLite 3.51.0 issue impacted the better-sqlite3 adapter. Prisma 7.3.0 bumps and pins the dependency used by @prisma/better-sqlite3 to prevent unexpected breakage. If you rely on this adapter, upgrading to 7.3.0 is strongly recommended.

Mapped enums behavior reverted for compatibility

In Prisma 7.0, mapped enums changed how values were generated when using @map, which turned out to be a sharp edge for many users migrating from Prisma 6. In 7.3.0, we’ve reverted that behavior to restore the prior, more compatible semantics.

BigInt precision preserved in JSON aggregation with relation joins

When using relationJoins with BigInt fields in Prisma 7, BigInt values could be returned as JSON numbers from PostgreSQL JSON aggregation, then silently lose precision when parsed in JavaScript (because values above Number.MAX_SAFE_INTEGER can’t be represented exactly).

In 7.3.0, BigInt fields are cast to text inside JSON aggregation, ensuring values come back as JSON strings and preserve full precision in JavaScript.

Upgrade notes

  • If you’re using @prisma/better-sqlite3, upgrading to 7.3.0 is recommended to avoid SQLite-related issues.
  • If you had to work around Prisma 7.0 mapped enum changes, those workarounds may no longer be necessary in 7.3.0.
  • Consider compilerBuild = "small" if bundle size matters more than raw performance (serverless/edge constraints), and stick with the default fast build for most production backends.

Be sure to follow us on social media to stay up to date with the latest releases of Prisma ORM and Prisma Postgres:

Don’t miss the next post!

Sign up for the Prisma Newsletter